Complex model development is inherently prone to error, requiring diligent review before reliance is placed on the outputs generated. We have experience providing model audits and reviews to corporates wishing to obtain third-party confirmation that the models they have developed are robust and free from material error.
The following types of review are available:
• Parallel build – the form of review with the most intensive resource requirement, parallel builds also give the highest form of assurance. Clients provide us with the inputs and ‘book of assumptions’ (i.e. documentation of the functionality, operation and output of the model to be rebuilt) and we develop a model to the stated specifications without the modeller having sighted the original client model. Once the parallel model is complete, its outputs are compared with those of the client model by a second member of our team, who reconciles the variances between the models, acting as a go-between between the respective model developers on our and our client’s team until all variances are resolved.
• Model audit – this form of review involves clients providing us with the model for review and its book of assumptions. We perform a series of procedures to determine that the model operates as set out in the book of assumptions, including by performing a review of the mathematical accuracy and appropriateness of each unique formula. We may, on request, also review the inputs used for appropriateness and transposition error (if compiled or incorporated from third sources).
• Model review – the form of review providing the least assurance, this approach does not review each unique formula. Rather, we consider the model from a high-level, strategic perspective, identifying whether it performs the intended calculations, whether those calculations are the appropriate ones for the strategic aim of the model and if there are omissions that should be addressed. We may also consider the inputs for appropriateness under this approach.